A turbine gas film hole surface roughness acquisition method based on non-destructive measurement

Abstract

The application discloses a turbine gas film hole surface roughness acquisition method based on non-destructive measurement, acquires a rough element wavelength, adaptively determines the weight length and point number of a noise reduction neighborhood in combination with the average point distance of an STL geometric file; searches a neighborhood point set through a KD-tree, constructs a local plane by using a Gaussian weight weighted singular value decomposition, and projects and denoises a target point, so that effective separation of overall geometric appearance and local roughness is realized; automatically identifies and removes gas film hole inlet and outlet edge vertices by calculating local plane fitting residuals, and eliminates measurement distortion caused by curvature mutation; the influence of uneven grid distribution is corrected by area weight, arithmetic average roughness, profile root mean square deviation and maximum profile height are calculated based on displacement before and after noise reduction; and the final result is output through iterative convergence judgment and inflection point truncation strategy. The application can realize accurate quantitative characterization of the inner surface roughness of the gas film hole under non-destructive conditions, and provides support for turbine blade processing qualification inspection, regular maintenance and cooling performance evaluation.

Description

Technical Field

[0001] This invention relates to the field of turbine mechanical geometric deviation detection and gas thermal performance evaluation technology, specifically a method for obtaining turbine film hole surface roughness based on non-destructive measurement. Background Technology

[0002] Compared to their design state, turbine blades inevitably suffer from structural deviations and damage after machining and service. Among these, incompletely exposed, small-scale structural deviations and damages that are highly sensitive to cooling performance are difficult to detect and even more difficult to quantitatively characterize. Taking film cooling holes as an example, their fully exposed area only includes the hole outlet and part of the windward side. During machining (such as femtosecond laser processing and EDM), painting, and service, significant roughness elements form on the inner wall of the film cooling hole. Furthermore, the film cooling hole itself possesses strong three-dimensional geometric features, further increasing the difficulty of detecting internal roughness. This significantly restricts the accurate assessment of the actual aerothermal performance of turbine blades. Turbine blade machining conformity inspection and periodic maintenance both require measurements to be conducted under non-destructive conditions, and determining whether actual performance meets standards also requires accurate acquisition of the corresponding damage structural parameters. Therefore, how to obtain the surface roughness of film cooling holes under non-destructive conditions has become the primary challenge in refined thermal analysis and conformity inspection.

[0003] Non-destructive testing of film cooling hole structure deviations and damage in turbine blades mainly includes methods such as surface profile measurement, CT scanning, and reverse reconstruction. For example, in "Geometric evolution and cooling performance degradation of film cooling holes under long-term ablation in advanced gasturbines" (Aerospace Science and Technology, 2026, Vol. 172, Article No. 111716), the authors obtained the morphology of the film cooling hole outlet by taking electron microscopic images of the surface of a simulated turbine blade ablation test specimen and using a hole edge profile image recognition algorithm. This method can identify the geometric features of the fully exposed area of ​​the film cooling hole relatively well, but it is difficult to obtain the structural information of the non-exposed area inside the hole, and it cannot achieve the detection and quantitative characterization of the roughness inside the hole. In the paper "Numerical investigation of the aerothermal mechanism induced by as-built topology inadditively manufactured film cooling holes" (Aerospace Science and Technology, 2026, Vol. 176, Article No. 112072), the authors used CT scanning combined with inverse reconstruction to obtain the full surface geometry of the air film pores. However, the root mean square roughness obtained in the scanned sample processing showed a negative value that contradicted the definition. The authors did not provide a reasonable explanation for this, nor did they provide a detailed process for roughness calculation.

[0004] To address the technical challenges of obtaining accurate quantitative assessment methods for deviations and damages in the film pore structure, it is urgent to propose a non-destructive measurement-based method for obtaining the surface roughness of turbine film pores, in order to support research on refined thermal analysis and structural conformity inspection of turbine machinery. Summary of the Invention

[0005] Technical problems to be solved

[0006] To overcome the shortcomings of existing technologies in the quantitative detection of the surface roughness of turbine film pores that are not fully exposed and have strong three-dimensional features, this invention proposes a method for obtaining the surface roughness of turbine film pores based on non-destructive measurement.

[0007] The core idea of ​​this invention is to determine the weight length and number of points of the noise reduction neighborhood based on the rough element wavelength and the average point spacing of the STL sheet, construct a local plane of the neighborhood point set of each target point and perform projection noise reduction, statistically analyze the target point offset before and after noise reduction, and calculate the roughness parameters. Simultaneously, by introducing area weights to correct uneven mesh distribution, introducing a boundary vertex protection mechanism to eliminate measurement distortion at the orifice edge, and ensuring appropriate noise reduction depth through iterative convergence judgment, this invention provides a precise and efficient roughness acquisition method for the qualification inspection of actual turbine film venting structures.

[0008] The technical solution of this invention is as follows:

[0009] The method for obtaining the surface roughness of turbine film pores based on non-destructive measurement includes the following steps:

[0010] Step 1: Import the STL geometry file of the air film aperture and calculate the average vertex distance of the triangular facet. ;

[0011] Step 2: Obtain the roughness element wavelength based on the inherent characteristic frequency band of the film atomization process. The weight length of the noise reduction neighborhood is calculated using the following formula. and points :

[0012]

[0013] in This is the scaling factor;

[0014] Step 3: For each target point Search its neighborhood using KD-tree The nearest points form a neighborhood set. ;

[0015] Step 4: Calculate the Gaussian weights of each neighboring point:

[0016]

[0017] in For the first Neighboring points and target point Spatial Euclidean distance; and construct weighted local center points. :

[0018]

[0019] Step 5: Calculate the unit normal vector of the local plane of the neighborhood point set using weighted singular value decomposition. ; target point Projecting along the normal vector onto this local plane, we obtain the denoised points:

[0020]

[0021] Step 6: Unify the direction of the normal vectors of all vertices so that they point inwards onto the surface: Calculate the center point of the STL geometry file after denoising. Calculate the decision parameters ,like Then flip the normal vector, if Then retain the normal vector;

[0022] Step 7: Calculate the area weight of each vertex. ;

[0023] Step 8: Calculate the local plane fitting residuals for each vertex:

[0024]

[0025] Will The vertices are marked as edge points, and the set of edge points is... Removed from roughness statistics, among which This is the edge sensitivity coefficient;

[0026] Step 9: Using the components of the target point displacement in the normal vector direction before and after noise reduction as the rough height, subtract the overall offset, and then apply this to the clean vertex set after removing edge vertices. Calculate the arithmetic mean roughness using the following formula. Root mean square deviation of profile and maximum profile height :

[0027]

[0028]

[0029]

[0030] in This is the precise value of the rough height after subtracting the overall offset; For the set of all vertices;

[0031] Step 10: Iteratively execute steps 3 through 9, for the... The next iteration is based on the set after removing edge vertices. Calculate the first Arithmetic mean roughness When satisfied

[0032]

[0033] When the convergence condition is met, the conservative estimate of the inflection point is used as the final output: If you want to skip to step 11, then skip to step 2. The STL geometry file after noise reduction is used as input for (3) to perform the denoising process. Sub-iteration noise reduction processing;

[0034] Step 11: Output the STL geometry file with roughness parameters and surface smoothing.

[0035] A further preferred embodiment is the roughness element wavelength described in step 2. The samples were obtained through non-destructive optical pre-scanning of the same batch of processing technology verification samples, or by querying the processing technology parameter archive database.

[0036] A further preferred embodiment is the average point distance mentioned in step 1. According to the formula

[0037]

[0038] Calculation, where This represents the total number of vertices in the STL geometry file. For the first The spatial Euclidean distance between a vertex and its next adjacent vertex in its storage order.

[0039] In a further preferred embodiment, the weighted singular value decomposition in step 5 is performed as follows:

[0040] Establish a decentralized matrix , ;

[0041] Construct a weighted diagonal matrix ;

[0042] Calculate the weighted covariance matrix ;

[0043] right Perform singular value decomposition ;

[0044] Take the right singular vector matrix The 3rd column vector in After normalization, the unit normal vector of the local plane is obtained. .

[0045] A further preferred option is the area weighting mentioned in step 7. Calculate as follows:

[0046] For each triangular facet Calculate its area:

[0047]

[0048] Add one-third of the area of ​​each triangular facet to its three vertices to obtain the area weight of each vertex:

[0049]

[0050] in Triangular facet The coordinates of the three vertices; For all target points The set of triangular faces with vertices.

[0051] A further preferred embodiment is the precise value of the roughness height mentioned in step 9. Calculate as follows:

[0052]

[0053]

[0054]

[0055] in, For single-point roughness height, This represents the overall normal offset of the STL geometry file caused by the noise reduction process.

[0056] In a further preferred embodiment, the roughness parameter in step 11 also includes a parameter relative to the radius of the air film pore. dimensionless value .

[0057] A further preferred embodiment is the set of edge points mentioned in step 8. This is an optional output of step 11, used to exclude edge distortion regions during subsequent refined thermal analysis.

[0058] In a further preferred embodiment, if the convergence condition is not met in step 10, the KD-tree search in step 3 is re-executed in each iteration to update the neighborhood topology relationship that has changed due to the change in vertex coordinates.

[0059] Beneficial effects

[0060] The present invention proposes a method for obtaining the surface roughness of turbine film pores based on non-destructive measurement, which has the following advantages compared with the prior art:

[0061] 1. This invention can obtain the surface roughness of the inner surface of the film film hole based on the STL geometric file of the film film hole without damaging the turbine blade and film film hole structure, thus meeting the non-destructive measurement requirements in machining qualification inspection and periodic maintenance.

[0062] 2. This invention can jointly determine the noise reduction neighborhood parameters based on the rough element wavelength (a prior process parameter) and the average point spacing of the STL sheet (data resolution), and achieve effective separation of the overall geometry of the film pores from local roughness undulations through Gaussian weighting, local plane construction, and projection noise reduction. This invention overcomes the blindness of general point cloud noise reduction methods that rely solely on pure geometric distance by using a process wavelength-driven kernel size reduction strategy.

[0063] 3. This invention corrects the effect of uneven patch distribution by using area weighting to obtain... , and By using roughness parameters, quantitative characterization of air film pore damage is achieved. At the same time, the distortion vertices at the inlet and outlet edges of the air film pores are automatically identified and removed through local fitting residuals to ensure the reliability of the statistical samples.

[0064] 4. This invention reduces single-processing errors by iteratively reducing noise and determining the convergence of roughness relative deviation results, providing support for refined thermal analysis, cooling performance evaluation, and structural conformity inspection of turbomachinery.

[0065] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0066] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which:

[0067] Figure 1 This is the complete process of surface roughness noise reduction in this embodiment of the invention;

[0068] Figure 2 This is a schematic diagram of the target points in the STL geometry file in this embodiment of the invention;

[0069] Figure 3 This refers to the neighborhood point set of the target point in the KD-tree search in this embodiment of the invention.

[0070] Figure 4 This refers to the local plane of the neighborhood point set of the target point in this embodiment of the invention.

[0071] Figure 5 In this embodiment of the invention, noise reduction is achieved by projecting the target point along the plane normal.

[0072] Figure 6 This is the STL geometry file for surface smoothing in this embodiment of the invention;

[0073] Figure 7 These are the air film pore roughness parameters calculated statistically in this embodiment of the invention. Detailed Implementation

[0074] The embodiments of the present invention are described in detail below. These embodiments are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0075] This invention proposes a method for obtaining the surface roughness of turbine film pores based on non-destructive measurement, comprising the following steps:

[0076] Step 1: Import the STL geometry file for the air film aperture and count the total number of vertices of all triangular facets in the file. Calculate the spatial Euclidean distance between adjacent vertices in the order they are stored in the file. Calculate the average vertex distance of the triangular facet using the following formula:

[0077]

[0078] in, The average point distance, This represents the total number of points in the STL geometry file. For the first The spatial Euclidean distance between a vertex and its next adjacent vertex in its storage order. In this embodiment, the average point distance of the STL geometry file It is 0.0123mm.

[0079] Step 2: Calculate the weighted length of the noise reduction neighborhood based on the rough element wavelength of the air film aperture cross-section. and points :

[0080]

[0081] in, and These represent the weight length and number of points of the noise reduction neighborhood, respectively. The wavelength of the rough element. The rough element is half the wavelength, i.e., the characteristic size. The scaling factor is set to 1.5 to 2.0 to ensure that the noise reduction neighborhood completely contains the single rough element.

[0082] The roughness element wavelength Based on the inherent characteristic frequency band of the film vacuum forming process: for femtosecond laser processing, The laser pulse energy density and scanning path spacing determine the process, which can be obtained through non-destructive optical pre-scanning of samples from the same batch of process verification parts or from a process archive database; for electrical discharge machining, It is determined by the discharge pulse energy and electrode feed parameters. As a priori input parameter for this method, its physical meaning is the average characteristic spacing of the roughness elements left on the inner wall of the hole by the machining process along the axial or circumferential direction of the hole.

[0083] In this embodiment, the roughness element half-wavelength The size is 0.124 mm, and the scale factor is [missing information]. If the value is 1.5, then the weight length of the noise reduction neighborhood is... It is 0.186mm, and the number of dots is... It is 150.

[0084] Step 3, KD-tree search Target points Within the neighborhood 1. Construct a neighborhood set of the target point using 1 point:

[0085] For each target point in the STL geometry file , The distance is searched in three-dimensional space using a KD-tree data structure. Recent The vertices constitute the neighborhood set of the target point:

[0086]

[0087]

[0088] in, The current target point for noise reduction is... The K-nearest neighbor algorithm is used. To the target point The closest 10 neighboring points (arranged in ascending order of distance).

[0089] In this step The value of is obtained by adaptive calculation based on rough element wavelength and average point distance in step 2, which ensures that the physical scale of the neighborhood coverage exactly contains a complete rough element feature, providing an accurate data basis for subsequent Gaussian weight calculation and local plane fitting.

[0090] In this embodiment, the target point distribution is as follows: Figure 2 As shown, the neighborhood point set obtained by KD-tree search is as follows: Figure 3 As shown.

[0091] Step 4, for the target point neighborhood point set The spatial Euclidean distance between each neighboring point and the target point is calculated sequentially, and weights are assigned according to the Gaussian kernel function. The closer the item, the greater its weight.

[0092]

[0093]

[0094] in The index number of the point in the neighborhood. For the first Neighboring points and target point Spatial Euclidean distance, For the first Gaussian weights of neighborhood points The weight length of the noise reduction neighborhood.

[0095] Step 5: Calculate the target point based on the obtained Gaussian weights and coordinates of the neighboring points. Weighted local center point coordinates within the neighborhood:

[0096]

[0097] in It is the weighted local center point (three-dimensional coordinate vector) of the neighborhood point set. The Gaussian weights calculated in step 4, Let be the three-dimensional coordinate vector of each neighboring point.

[0098] Singular value decomposition calculates the normal vector of the local plane of the neighborhood point set:

[0099] The neighborhood points are decentered, and singular value decomposition is used to calculate the tangent and normal vectors of the local plane of the neighborhood point set, and the normal vectors are normalized.

[0100]

[0101]

[0102]

[0103]

[0104] in, For the first Neighboring points Relative to the weighted local center point directional vector, The direction vectors of all neighboring points As a decentralized matrix formed by concatenating column vectors, that is ; For distributing Gaussian weights only along the main diagonal phalanx; For target point The second-order moment matrix of the neighborhood point set is used to quantitatively characterize the anisotropic discrete distribution characteristics of the neighborhood point set in three-dimensional space. To The column vectors of the left singular vector matrix obtained after singular value decomposition form an orthonormal basis in the three-dimensional space of the neighborhood point set. This is the singular value diagonal matrix obtained after singular value decomposition; This is the transpose of the right singular vector matrix after singular value decomposition; A right singular vector matrix The third column vector in the vector corresponds to the minimum singular value; Let be the unit normal vector of the local plane of the neighborhood point set.

[0105] Build target points Local plane of the neighborhood point set:

[0106] Weighted local center point Let be a point on the plane, with the unit normal vector Construct the target point using the plane normal vector. Local plane of the neighborhood point set:

[0107]

[0108] in Let be the free variables in three-dimensional space (i.e., the coordinates of any point on this local plane). The equation defines the free variables passing through the point . And with The plane is the normal vector. In this embodiment, the local plane of the neighborhood point set is constructed as follows: Figure 4 As shown.

[0109] Target point projection noise reduction:

[0110] target point Along the unit normal vector Projecting the direction onto the local plane of the neighborhood point set yields the new vertex coordinates after noise reduction:

[0111]

[0112] in Represents the original target point The orthogonal projection point on the local plane of the neighborhood point set is the new vertex after noise reduction. The geometric meaning of this formula is: [The original point...] unit normal vector along the local plane Translate in direction, and the translation distance is The perpendicular distance to the plane, thus falling to the foot of the perpendicular on the plane. .

[0113] After processing all triangle facet vertices in the STL geometry file using steps 3 to 5, this round of surface roughness denoising is complete. In this embodiment, the process of denoising by projecting the target point along the plane normal is as follows: Figure 5 As shown.

[0114] Step 6: Unify the direction of the normal vector:

[0115] Calculate the center point coordinates of the STL geometry file after surface roughness denoising, and use the direction pointing to the inside of the surface as the positive direction to unify the direction of the unit normal vector of the local plane of the neighborhood point set;

[0116] First, calculate the geometric center of all vertices after noise reduction. :

[0117]

[0118] For each target point Calculate from the geometric center point Pointing to the point after noise reduction The signed distance of the vector in the direction of the normal vector , which serves as a parameter for determining the direction of the normal vector:

[0119]

[0120] in It is a scalar. If ,illustrate Located at the geometric center of In the positive direction, i.e., the current normal vector Pointing to the outside of the curved surface (away from the interior of the solid), it should be flipped; if This indicates that the current normal vector points inside the surface (inside the solid) and should be retained. The rules for unifying the direction of normal vectors are as follows:

[0121]

[0122] in The unit normal vector of the local plane of the neighborhood point set after direction unification always points inward to the surface.

[0123] Step 7: Calculate the area weights of the vertices:

[0124] Calculate the area weights of the vertices of triangular facets in an STL geometry file to remove the influence of local density distribution of facets on roughness statistics:

[0125]

[0126]

[0127] in, Triangular facet The area; Triangular facet The coordinates of the three vertices; For all target points The set of triangular faces with vertices; For target point The equivalent area weight.

[0128] Step 8: Detect and mark the vertices of the inlet and outlet edges of the air film orifice, and remove them from the roughness statistical sample:

[0129] For each target point, calculate the local plane fitting residual of its neighborhood point set:

[0130]

[0131] like , The edge sensitivity coefficient, ranging from 3.0 to 5.0, indicates that the target point is located in the edge region of the air film vent inlet / outlet and is marked as the target point. , Let be the set of edge points.

[0132] Step 9: Calculate the precise value of the roughness height:

[0133] Target points before and after surface roughness noise reduction The component of the displacement in the normal direction is used as the roughness height, and the overall offset of the STL geometry file caused by the noise reduction process is subtracted to obtain the accurate value of the roughness height.

[0134]

[0135]

[0136]

[0137] in, For single-point roughness height, This represents the overall normal offset of the STL geometry file caused by the noise reduction process. This is the precise value of the rough height after subtracting the overall offset; It is the set of all vertices.

[0138] Calculate the arithmetic mean roughness:

[0139] Based on the precise value of the roughness height, the arithmetic mean roughness obtained after surface roughness denoising is calculated:

[0140]

[0141] In this embodiment, the arithmetic mean roughness is obtained after the first surface noise reduction. It is 0.002588.

[0142] Calculate the root mean square deviation of the profile:

[0143] Based on the precise value of the roughness height, the root mean square deviation of the profile obtained after surface roughness denoising is calculated:

[0144]

[0145] In this embodiment, the root mean square deviation of the contour is obtained after the first surface noise reduction. It is 0.003421.

[0146] Calculate the maximum profile height:

[0147] Based on the precise value of the roughness height, the maximum contour height obtained after surface roughness denoising is calculated:

[0148]

[0149] The maximum contour height obtained after the first surface noise reduction in this embodiment It is 0.044662.

[0150] Step 10: Iterate through steps 3 through 9, and set the surface roughness noise reduction test criteria and inflection point cutoff strategy:

[0151] For the Next iteration of noise reduction , Let the number of noise reduction iterations be defined as:

[0152]

[0153]

[0154]

[0155]

[0156] in For the first After the second noise reduction process, as the first The target point's three-dimensional coordinate vector is input for noise reduction; For the first After the noise reduction process, the target point is projected onto the local plane as a new three-dimensional coordinate vector, which is the output point of this round of noise reduction. For the first In the noise reduction process, the unit normal vector of the local plane of the neighborhood point set of the target point is... For the first In the noise reduction process, the single-point roughness height of the target point is... For the first After the second noise reduction process, the area weight of the target point. For the first The overall normal offset of the STL geometry file caused by the noise reduction. For the first In the noise reduction process, the rough height is the precise value after subtracting the overall offset from the target point. For the first The arithmetic mean roughness obtained after the noise reduction process;

[0157] Set convergence criteria:

[0158]

[0159] If the above conditions are met, then the first... The noise reduction process has entered the risk zone of over-smoothing, so a conservative estimate of the inflection point is used as the final output: Then proceed to step 11; if the convergence condition is not met, then because the vertex coordinates have been updated after noise reduction, the spatial neighborhood topology has changed, and the vertex will be... The STL geometry file after noise reduction is used as input for step 3, and then... This is the second iteration of noise reduction processing. The initial condition for the iteration is: hour, The vertex coordinates of the original input STL geometry file; hour, These are the vertex coordinates after the previous noise reduction process.

[0160] In this embodiment, the arithmetic mean roughness obtained in the third surface denoising process has a relative deviation of 7.74% from that in the second process. Therefore, the STL geometric file obtained in the first surface denoising process has reached the surface smoothness standard, and the roughness parameter of the first process is the final value.

[0161] Step 11: Output roughness parameters and boundary vertex markers:

[0162] After the surface roughness and noise reduction test standard is passed, the roughness parameters are output. , and and its relative to the film pore radius dimensionless value STL geometry file after surface smoothing, set of boundary vertex markers This is used to exclude edge distortion areas during subsequent refined thermal analysis. For example... Figure 6 and Figure 7In this embodiment, the radius of the air film hole is 0.25 mm, the relative value of the radius of the arithmetic mean roughness is 0.01035, the relative value of the radius of the root mean square deviation of the profile is 0.01368, and the relative value of the radius of the maximum profile height is 0.17865.

[0163] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention without departing from the principles and spirit of the present invention.

Claims

1. A method for obtaining the surface roughness of turbine film pores based on non-destructive measurement, characterized in that: Includes the following steps: Step 1: Import the STL geometry file of the air film aperture and calculate the average vertex distance of the triangular facet. ; Step 2: Obtain the roughness element wavelength based on the inherent characteristic frequency band of the film atomization process. The weight length of the noise reduction neighborhood is calculated using the following formula. and points : in This is the scaling factor; Step 3: For each target point Search its neighborhood using KD-tree The nearest points form a neighborhood set. ; Step 4: Calculate the Gaussian weights of each neighboring point: in For the first Neighboring points and target point Spatial Euclidean distance; and construct weighted local center points. : Step 5: Calculate the unit normal vector of the local plane of the neighborhood point set using weighted singular value decomposition. ; target point Projecting along the normal vector onto this local plane, we obtain the denoised points: Step 6: Unify the direction of the normal vectors of all vertices so that they point inwards onto the surface: Calculate the center point of the STL geometry file after denoising. Calculate the decision parameters ,like Then flip the normal vector, if Then retain the normal vector; Step 7: Calculate the area weight of each vertex. ; Step 8: Calculate the local plane fitting residuals for each vertex: Will The vertices are marked as edge points, and the set of edge points is... Removed from roughness statistics, among which This is the edge sensitivity coefficient; Step 9: Using the components of the target point displacement in the normal vector direction before and after noise reduction as the rough height, subtract the overall offset, and then apply this to the clean vertex set after removing edge vertices. Calculate the arithmetic mean roughness using the following formula. Root mean square deviation of profile and maximum profile height : in This is the precise value of the rough height after subtracting the overall offset; For the set of all vertices; Step 10: Iteratively execute steps 3 through 9, for the... The next iteration is based on the set after removing edge vertices. Calculate the first Arithmetic mean roughness When satisfied When the convergence condition is met, the conservative estimate of the inflection point is used as the final output: If you want to skip to step 11, then skip to step 2. The STL geometry file after noise reduction is used as input for step 3, and then... Sub-iteration noise reduction processing; Step 11: Output the STL geometry file with roughness parameters and surface smoothing.

2. The method for obtaining turbine film pore surface roughness based on non-destructive measurement according to claim 1, characterized in that: The roughness element wavelength mentioned in step 2 The samples were obtained through non-destructive optical pre-scanning of the same batch of processing technology verification samples, or by querying the processing technology parameter archive database.

3. The method for obtaining turbine film pore surface roughness based on non-destructive measurement according to claim 1, characterized in that: The average point distance mentioned in step 1 According to the formula Calculation, where This represents the total number of vertices in the STL geometry file. For the first The spatial Euclidean distance between a vertex and its next adjacent vertex in its storage order.

4. The method for obtaining turbine film pore surface roughness based on non-destructive measurement according to claim 1, characterized in that: The weighted singular value decomposition described in step 5 is performed as follows: Establish a decentralized matrix , ; Construct a weighted diagonal matrix ; Calculate the weighted covariance matrix ; right Perform singular value decomposition ; Take the right singular vector matrix The 3rd column vector in After normalization, the unit normal vector of the local plane is obtained. .

5. The method for obtaining turbine film pore surface roughness based on non-destructive measurement according to claim 1, characterized in that: The area weighting mentioned in step 7 Calculate as follows: For each triangular facet Calculate its area: Add one-third of the area of ​​each triangular facet to its three vertices to obtain the area weight of each vertex: in Triangular facet The coordinates of the three vertices; For all target points The set of triangular faces with vertices.

6. The method for obtaining turbine film pore surface roughness based on non-destructive measurement according to claim 1, characterized in that: The precise value of roughness height mentioned in step 9 Calculate as follows: in, For single-point roughness height, This represents the overall normal offset of the STL geometry file caused by the noise reduction process.

7. The method for obtaining turbine film pore surface roughness based on non-destructive measurement according to claim 1, characterized in that: The roughness parameter mentioned in step 11 also includes the roughness parameter relative to the radius of the air film pore. dimensionless value .

8. The method for obtaining turbine film pore surface roughness based on non-destructive measurement according to claim 1, characterized in that: The set of edge points mentioned in step 8 This is an optional output of step 11, used to exclude edge distortion regions during subsequent refined thermal analysis.

9. The method for obtaining turbine film pore surface roughness based on non-destructive measurement according to claim 1, characterized in that: In step 10, if the convergence condition is not met, the KD-tree search in step 3 is re-executed in each iteration to update the neighborhood topology relationship that has changed due to the change in vertex coordinates.

Images